Functional traits mediate individualistic species-environment distributions at broad spatial scales while fine-scale species associations remain unpredictable

Abstract

Premise: Numerous processes influence plant distributions and co-occurrence patterns, including ecological sorting, limiting similarity, and stochastic effects. To discriminate among these processes and determine the spatial scales at which they operate, we investigated how functional traits and phylogenetic relatedness influence the distribution of temperate forest herbs.

Methods: We surveyed understory plant communities across 257 forest stands in Wisconsin and Michigan (USA) and applied Bayesian phylogenetic linear mixed-effects models (PGLMMs) to quantify how functional traits and phylogenetic relatedness influence the environmental distribution of 139 herbaceous plant species along broad edaphic, climatic, and light gradients. These models also allowed us to test how functional and phylogenetic similarity affect species co-occurrence within microsites.

Results: Leaf height, specific leaf area, and seed mass all influenced individualistic plant distributions along landscape-scale gradients in soil texture, soil fertility, light availability, and climate. In contrast, phylogenetic relationships did not consistently predict species-environment relationships. Neither functionally similar nor phylogenetically related herbs segregated among microsites within forest stands.

Conclusions: Trait-mediated ecological sorting appears to drive temperate-forest community assembly, generating individualistic plant distributions along regional environmental gradients. This finding links classic studies in plant ecology and prior research in plant physiological ecology to current trait-based approaches in community ecology. However, our results fail to support the common assumption that limiting similarity governs local plant co-occurrences. Strong ecological sorting among forest stands coupled with stochastic fine-scale interactions among species appear to weaken deterministic, niche-based assembly processes at local scales.

Keywords: Wisconsin; community assembly; environmental filtering; functional traits; limiting similarity; phylogeny; temperate forest.

Figure 1

Map of study sites in Wisconsin, USA. Vegetation surveys were conducted at 257 forest stands across Wisconsin between 2000 and 2012. Community classification follows Curtis (1959): Northern Upland Forests (NUF), Pine Barrens (PB), Southern Lowland Forests (SLF), and Southern Upland Forests (SUF). The gray band illustrates the location of a pronounced floristic tension zone marking the range of limits of plant species with northern vs. southern affinities (see Curtis, 1959).

Figure 2

Heatmap illustrating distributional and co‐occurrence pattern among 139 herbaceous species (rows) distributed across 257 study sites (columns) in Wisconsin, USA. Filled cells represent sites in which a particular species was present. The tree illustrates phylogenetic relationships among herbaceous plant species. Black bars and labels correspond to community types (Curtis, 1959): SUF = Southern Upland Forests, NUF = Northern Upland Forests, PB = Pine Barrens, SLF = Southern Lowland Forests (for description of community types, see text).

Figure 3

Estimated coefficients from PGLMM‐1 quantifying the effects of trait and environmental predictors on the probability of species occurrence. Top panel depicts main effects of predictors; bottom four panels depict estimates for trait‐environment interactions. Coefficients greater than zero indicate a positive relationship between predictors and the probability of occurrence. Solid circles indicate estimates that are different from zero (95% credible interval does not overlap with zero).

Figure 4

Graphical depiction of (A–D) species‐environment distributions and (E–J) the six important trait‐environment interactions shown in Figure 1. In A–D, lines illustrate how the probability of species occurrence varies with soil fertility (A), soil texture (B), tree basal area (C), and the ratio of precipitation to potential evapotranspiration (D) for a representative sample of 20 herb species (see Appendices [Link], [Link], [Link], [Link] for all 139 herb species). In E–J, lines depict the predicted environmental distribution of herb species with low, intermediate, and high trait values (corresponding to the 5%, 50%, and 95% quantiles of herb trait distributions). These marginal predictions from PGLMM‐1 were calculated holding all other trait and environmental effects constant. Note that individualistic species‐environment relationships (A–D) emerge from the additive effects of traits, environmental conditions, trait‐environment interactions, and residual variation in how species respond to the environment.

Figure 5

Patterns of within‐site (A) functional repulsion/attraction and (B) phylogenetic repulsion from/attraction to PGLMM‐3 (Appendix S12). These histograms display ΔWAIC values; differences in WAIC values between PGLMMs with and without covariance matrices specify quadrat‐scale repulsion (top) or attraction (bottom) for all 257 forest sites. Only ΔWAIC values >5 (vertical dashed line) provide strong evidence of functional/phylogenetic repulsion (spatial segregation of functionally similar or phylogenetically related species among microsites within forest stands) or functional/phylogenetic attraction (spatial aggregation within microsites). Negative ΔWAIC values reflect cases where the penalty for additional complexity is greater than the amount of additional information gained by accounting for functional or phylogenetic repulsion.